*Check out the full Two by Two Playbook here.*

Original article

The AI Plateau Is Real — How We Jump To The Next Breakthrough

Key Takeaway

• For entrepreneurs: The next breakthrough in AI will come from leveraging high-quality, proprietary business data, which presents opportunities for startups to develop novel ways to source, prepare, and utilize this data.
• For investors: Investing in startups that focus on unlocking and utilizing business data for AI training could yield significant returns, as this data is more valuable and less exploited than public internet data.

Summary

The article discusses the current plateau in AI development, comparing it to historical technological advancements that follow an S-Curve pattern of rapid innovation followed by stabilization. The authors argue that the next leap in AI performance will come from utilizing proprietary business data, which is of higher quality and more abundant than the public data currently used. They identify four key areas of opportunity for startups: engaging experts for high-quality data, leveraging latent data within business apps, capturing data in context without disrupting workflows, and securing proprietary data by helping enterprises build their own custom models.

Insights

• Historical Context: Technological advancements often follow an S-Curve pattern, with rapid innovation followed by stabilization. Examples include TCP/IP, browser technology, and mobile apps.
• AI Plateau: The current AI models, despite initial rapid improvements, are now showing limited incremental progress. This is due to the exhaustion of public data sources and the need for higher-quality data.
• Proprietary Business Data: Data produced in a work context is more valuable and abundant than public data. Examples include product specs, sales decks, medical studies, and data from platforms like Zoom, Slack, and Notion.
• Opportunities for Startups:
  1. Engage Experts: Source high-quality data from experts in each field, using novel incentive structures and community engagement.
  2. Leverage Latent Data: Help enterprises prepare and utilize data from business apps for AI training.
  3. Capture in Context: Develop methods to capture new data without disrupting workflows, including multimodal content.
  4. Secure the Secret Sauce: Assist enterprises in creating and deploying their own custom models to protect proprietary IP.

Implications

• Data Quality: High-quality data is crucial for the next breakthrough in AI, and this data is predominantly found in proprietary business contexts.
• Privacy and Security: Enterprises need to protect their proprietary data and ensure that AI models are aligned with their goals and values.
• Investment Opportunities: Startups that address these challenges could create significant value and attract substantial investment.
• Human-Centric AI: The next wave of AI should prioritize human discovery, knowledge, privacy, and quality.